Caelan Gander

Forward-deployed with the client

I was embedded with the client, running their working sessions and carrying their decisions into engineering as the point of contact between the client and the build. Where a requirement was ambiguous, I increasingly built a small prototype with AI to settle it before writing the ticket. I also owned two-plus years of release communications, every release note and user-guide update, including prospect-facing docs leadership called “incredibly helpful.”

The systems that were genuinely hard

• Multi-level, hour-based authorization engine. Total-hour limits sit above per-week, which sit above per-day, with hours that can be shared across a group of clients. Getting the precedence and the draw-down order right is what decides whether a visit is authorized, and whether it can be billed.
• Tiered, configurable reimbursement engine. Rates vary by program, payor, and service line, so the engine had to be data-driven rather than hard-coded, with the tiers and contract variables exposed as configuration the client could reason about.
• Authorization-sync API. I updated the bi-directional, real-time API that keeps authorizations in lockstep between the two systems – specifying the field mappings, the sync triggers, and the conflict behavior, including mapping each discipline to the right procedure codes.

Tracking the money, bill to reimbursement

• End-to-end authorization tracking. Complex authorizations have to be followed from the bill all the way to the payment. I built the tracking that keeps each claim accurate as charges roll up into a single code and bill out as timed units, so what's owed is what gets billed and reimbursed.
• Two reimbursement-review reports, built from scratch with the client. Each compares expected reimbursement against what actually landed on the claim. They grew from 30 to more than 70 columns – every one a claim-relevant field or value, several calculated from existing data to surface insight the client didn't have before. Operations and billing review differently, so I split the work into an hourly authorization report for operations and a claim review report for billing, each with its own submission criteria.
• A charge inquiry window I designed. Sort and pick a bill, and a pop-out shows exactly which charges hit it and traces each one back to the General Ledger.

And the unglamorous-but-critical part

I owned testing readiness for the cutover, migrating the automated test suite from Flux to Playwright so go-live wasn't flying blind.